Analyzing BANKSY Embedding for Cell Identities

Introduction

In this doc, we will be looking at the banksy clusters and try to interpret them based on their identity, and other characteristics.

Load the data

The objects are here:

1. RDS object with Leiden Clusters: /xchip/beroukhimlab/coja/Spatial_PLGG/data/Xenium/Xenium_Objects/total_spatial_banksy_clusters_20240711_135041.rds

   • /xchip/beroukhimlab/coja/Spatial_PLGG/data/Xenium/Xenium_Objects/total_spatial_banksy_embeddings/total_spatial_banksy_clusters_subset_20240711_135041.rds

2. RDS object with cell type cluster markers: /xchip/beroukhimlab/coja/Spatial_PLGG/data/Xenium/Xenium_Objects/total_spatial_banksy_clusters_markers_20240711_135041.rds

Metadata is here:

1. /Users/youyun/Documents/HMS/PhD/beroukhimlab/broad_mount/coja/Spatial_PLGG/data/metadata/Xenium_PS.xlsx

banksy_embeddings = readRDS(paste0(
    workdir,'coja/Spatial_PLGG/data/Xenium/Xenium_Objects/total_spatial_banksy_clusters_subset_20240719_092550.rds'
))
colData(banksy_embeddings)$sample_id = factor(colData(banksy_embeddings)$sample_id)

cell_type_markers = fread(paste0(
    workdir,'coja/Spatial_PLGG/data/markers/final_manual_markers.csv'
))
xenium_markers = fread(paste0(
    workdir, 'youyun/plgg/data/xenium_selection/Xenium_hBrain_v1_metadata.csv'
))

Looking at the harmony corrected embeddings

We will first look at the extent of batch effect in the cohort by looking at the embeddings before and after harmony correction.

plot_grid(
    ggdraw() + draw_label("BANKSY Embedding UMAP Before/After Harmony (Cell Identity)", fontface='bold'),
    plot_grid(
        plotReducedDim(
            banksy_embeddings, "UMAP_M1_lam0.2", point_size = 0.1,
            point_alpha = 0.5, color_by = "sample_id"
        ) +
            theme(legend.position = "none"),
        plotReducedDim(
            banksy_embeddings, "UMAP_Harmony_BANKSY_lam0.2", 
            point_size = 0.1,point_alpha = 0.5, color_by = "sample_id"
        ) +
            theme(legend.title = element_blank()) +
            guides(colour = guide_legend(override.aes = list(size = 5, alpha = 1))),
        nrow = 1,
        rel_widths = c(1, 1.2)
    ), ncol = 1,rel_heights = c(0.1, 5)
)

plot_grid(
    ggdraw() + draw_label("BANKSY Embedding UMAP Before/After Harmony (Neighborhood)", fontface='bold'),
    plot_grid(
        plotReducedDim(
            banksy_embeddings, "UMAP_M1_lam0.8", point_size = 0.1,
            point_alpha = 0.5, color_by = "sample_id"
        ) +
            theme(legend.position = "none"),
        plotReducedDim(
            banksy_embeddings, "UMAP_Harmony_BANKSY_lam0.8", 
            point_size = 0.1,point_alpha = 0.5, color_by = "sample_id"
        ) +
            theme(legend.title = element_blank()) +
            guides(colour = guide_legend(override.aes = list(size = 5, alpha = 1))),
        nrow = 1,
        rel_widths = c(1, 1.2)
    ), ncol = 1,rel_heights = c(0.1, 5)
)

What do the cell clusters look like?

res = 1
clust_name = gsub('.0$','',sprintf('clust_Harmony_BANKSY_lam0.2_k50_res%.1f',res))

We will be looking at resolution = 1 for the clusters using the clust_Harmony_BANKSY_lam0.2_k50_res1 column.

plot_grid(
    ggdraw() + draw_label("BANKSY Embedding UMAP for Cell Identity(L) and Niche(R)", fontface='bold'),
    plot_grid(
        plotReducedDim(
            banksy_embeddings, "UMAP_Harmony_BANKSY_lam0.2", 
            point_size = 0.1, point_alpha = 0.5, 
            color_by = clust_name
        ) +
            theme(legend.title = element_blank()) +
            guides(colour = guide_legend(override.aes = list(size = 5, alpha = 1))),
        plotReducedDim(
            banksy_embeddings, "UMAP_Harmony_BANKSY_lam0.8", 
            point_size = 0.1, point_alpha = 0.5, 
            color_by = gsub('lam0.2','lam0.8',clust_name)
        ) +
            theme(legend.title = element_blank()) +
            guides(colour = guide_legend(override.aes = list(size = 5, alpha = 1))),
        nrow = 1,
        rel_widths = c(1, 1)
    ), ncol = 1,rel_heights = c(0.1, 5)
)

Just to get an idea of where each cluster is, we can look at the UMAPs colored individually by the clusters.

cell_clusters = sort(unique(as.numeric(colData(banksy_embeddings)[,clust_name])))
# one hot encoded columns
one_hot_encoded_clusters = data.frame(dcast(data.table(
    cell_sample_id = rownames(colData(banksy_embeddings)),
    clusters = paste0('cell_in_',colData(banksy_embeddings)[,clust_name])
), cell_sample_id ~ clusters, length)[order(match(cell_sample_id,rownames(colData(banksy_embeddings))))])
# cbind into the colData
if(all(rownames(colData(banksy_embeddings)) == colData(banksy_embeddings)$cell_sample_id)){
    colData(banksy_embeddings) = cbind(
        colData(banksy_embeddings),
        one_hot_encoded_clusters
    )
}else{
    stop('Cell IDs do not match')
}

# make the plots
a = lapply(cell_clusters, function(x){
    assign(
        paste0('cell_type_plot_',x),
        plotReducedDim(
            banksy_embeddings, "UMAP_Harmony_BANKSY_lam0.2", 
            point_size = 0.1, point_alpha = 0.5, 
            color_by = paste0('cell_in_',x)
        )  + theme(
            legend.position = "none",
            axis.text=element_blank(),
            axis.ticks=element_blank(),
            axis.title=element_blank()
        ) + labs(title = paste0('Cluster ',x)),
        envir = .GlobalEnv
    )
})

plot_grid(
    ggdraw() + draw_label("BANKSY Embedding UMAP by Cluster", fontface='bold'),
    plot_grid(
        plotlist = mget(paste0('cell_type_plot_',cell_clusters)),
        ncol = 6,
        rel_widths = rep(1,length(cell_clusters))
    ), ncol = 1,rel_heights = c(0.1, 5)
)

What are the cell types in the clusters?

Manually- and Xenium- annotated cell type markers

markers_to_score = rbind(cell_type_markers[
    category == 'Cell Type' & xenium == TRUE,
    .(markers = marker),.(annotation = paste0(annotation, '_manual'))
],xenium_markers[
    ,.(markers = Genes),.(annotation = paste0(Annotation, '_xenium'))
])[
    ,annotation := make.names(annotation)
]
markers_to_score = markers_to_score[annotation %in% markers_to_score[,.N, .(annotation)][N>=2]$annotation]

# Scoring the cells by cell type markers
cell_types = unique(markers_to_score$annotation)
marker_score = do.call('cbind',lapply(cell_types, function(x){
    genes_of_interest = markers_to_score[annotation == x]$markers
    cell_scores = log10(colSums(
        assay(banksy_embeddings[genes_of_interest,], 'normcounts')
    )/length(genes_of_interest) + 1)
    cell_scores_df = data.frame(cell_scores)
    rownames(cell_scores_df) = names(cell_scores)
    colnames(cell_scores_df) = x
    cell_scores_df
}))
colData(banksy_embeddings)[,cell_types] = NULL
if(all(rownames(colData(banksy_embeddings)) == rownames(colData(banksy_embeddings)))){
    colData(banksy_embeddings) = cbind(colData(banksy_embeddings),marker_score)
}else{
    stop('Cell IDs do not match')
}

# make the plots
a = lapply(cell_types, function(x){
    assign(
        paste0('plot_',x),
        plotReducedDim(
            banksy_embeddings, "UMAP_Harmony_BANKSY_lam0.2", 
            point_size = 0.1, point_alpha = 0.5, 
            color_by = x
        ) +
            theme(legend.title = element_blank(), legend.position = 'bottom') +
            guides(colour = guide_legend(override.aes = list(size = 5, alpha = 1), nrow = 1)) +
            labs(title = x),
        envir = .GlobalEnv
    )
})

We are going to annotate 34 cell types. and we will now visualize the marker scores.

plot_grid(
    ggdraw() + draw_label("BANKSY Embedding UMAP by Manually Curated Markers", fontface='bold'),
    plot_grid(
        plotlist = mget(paste0('plot_',grep('_manual',cell_types, value = TRUE))),
        ncol = 4,
        rel_widths = rep(1,sum(grepl('_manual',cell_types)))
    ), ncol = 1,rel_heights = c(0.1, 5)
)

plot_grid(
    ggdraw() + draw_label("BANKSY Embedding UMAP by Xenium Markers", fontface='bold'),
    plot_grid(
        plotlist = mget(paste0('plot_',grep('_xenium',cell_types, value = TRUE))),
        ncol = 5,
        rel_widths = rep(1,sum(grepl('_xenium',cell_types)))
    ), ncol = 1,rel_heights = c(0.1, 5)
)

We should complement the manual approach with more statistics

Looking at these ‘metagenes’ statistically

A better way to do this is to test statistically, which clusters are enriched for which signature. We will do that here

cell_type_cluster_sig = do.call('rbind',lapply(paste0('cell_in_',cell_clusters), function(x){
    scran_cell_markers = findMarkers(
        as.matrix(t(colData(banksy_embeddings)[,cell_types])),
        groups = colData(banksy_embeddings)[,x],
        test.type="wilcox", direction="up"
    )    
    delta_dt = melt(setDT(
        as.data.frame(colData(banksy_embeddings)[,c(cell_types,x)]),
        keep.rownames = 'cell_sample_id'
    ), id.vars = c('cell_sample_id',x), variable.name = 'cell_type', value.name = 'score')[
        ,.(delta_score = mean(score[get(x) == 1]) - mean(score[get(x) == 0])),.(cell_type)
    ]
    cluster_cell_type = setDT(
        data.frame(scran_cell_markers[[2]]),
        keep.rownames = 'cell_type'
    )[,.(cell_type = make.names(cell_type), Top, p.value, FDR, cluster = x)]
    merge(
        cluster_cell_type, delta_dt, by = 'cell_type'
    )
}))

# make a hierarchically clustered heatmap
cell_type_cluster_sig_matrix = dcast(
    cell_type_cluster_sig[,.(cell_type, cluster, FDR = ifelse(FDR<0.05,'*',''))],
    cluster ~ cell_type, value.var = 'FDR'
)
cell_type_cluster_delta_matrix = dcast(
    cell_type_cluster_sig[,.(cell_type, cluster, delta_score)],
    cluster ~ cell_type, value.var = 'delta_score'
)
plot_sig_mt = as.matrix(cell_type_cluster_sig_matrix[,c(2:ncol(cell_type_cluster_sig_matrix)), with = FALSE])
plot_delta_mt = as.matrix(cell_type_cluster_delta_matrix[,c(2:ncol(cell_type_cluster_delta_matrix)), with = FALSE])
rownames(plot_sig_mt) = cell_type_cluster_sig_matrix$cluster
rownames(plot_delta_mt) = cell_type_cluster_delta_matrix$cluster
col_fun = colorRamp2(c(-0.3,0,0.3), c("red", "white", "green"))
Heatmap(
    plot_delta_mt, name = "Delta Mean Norm Counts", col = col_fun,
    cell_fun = function(j, i, x, y, width, height, fill) {
        grid.text(sprintf("%s", plot_sig_mt[i, j]), x, y, gp = gpar(fontsize = 10))
})

Looking at the DE Genes in each cluster

ggplot(colData(banksy_embeddings)) + 
    geom_bar(aes(x = get(clust_name), fill = as.numeric(get(clust_name)))) + 
    facet_wrap(~sample_id, nrow = 3, scale = 'free_y') +
    theme(
        axis.text.x = element_text(angle = 45, hjust = 1),
        text = element_text(size = 15)
    ) + labs(
        x = 'Cluster ID', y = 'Number of Cells (log10)'
    ) + scale_fill_viridis_c(option = 'D') +
    guides(fill = guide_legend(title = 'Cluster ID')) +
    scale_y_log10()

banksy_embeddings_seurat = as.Seurat(banksy_embeddings, counts = 'counts', data = NULL)
banksy_embeddings_seurat = NormalizeData(
    banksy_embeddings_seurat, normalization.method = "LogNormalize", 
    scale.factor = 10000
)
Idents(banksy_embeddings_seurat) = clust_name

DE_genes_wilcox = data.table(FindAllMarkers(
    banksy_embeddings_seurat,
    slot = 'counts',
    test.use = 'wilcox',
    only.pos = TRUE,
    verbose = FALSE
))

DE_genes_wilcox = merge(
    DE_genes_wilcox,
    xenium_markers[,.(gene = Genes, Annotation)],
    by = 'gene'
)

kable(DE_genes_wilcox[order(p_val_adj)][
    p_val_adj <= 0.05 & avg_log2FC >= 1, .(
        N = .N, 
        `Markers` = paste0(
            gene[avg_log2FC >= 1], ' (', 
            round(avg_log2FC[avg_log2FC >= 1], 2),  ', ',
            Annotation[avg_log2FC >= 1] ,
            ')', collapse = ', '
        )
    ), 
    .(cluster)
][order(cluster)])

cluster	N	Markers
1	14	ABCC9 (1.33, VLMC), B4GALNT1 (1.32, Glioblastoma (Cancer cells)), CNTN2 (1.43, Oligodendrocyte), FGFR3 (1.02, Astrocyte), GAD2 (1.03, Lamp5 Lhx6), IGFBP5 (1.1, Glioblastoma (Cancer cells)), NR4A2 (1.79, Glioblastoma (TME)), PLD5 (1.27, Pax6), RELN (1.47, Pax6), SDK1 (1.8, Chandelier), SPHKAP (1.31, Lamp5 Lhx6), TRAC (1.14, Glioblastoma (TME)), TRPC6 (1.39, Sst Chodl), UNC5B (1.14, Chandelier)
2	12	FBLN1 (1.69, VLMC), NNAT (3.09, Glioblastoma (Cancer cells)), SOX11 (2.69, Glioblastoma (Cancer cells)), CUX2 (2.18, L2/3 IT), HS3ST4 (2.34, L6 CT), LRRK2 (2.47, OPC), NOTCH1 (1.58, Endothelial), RASGRP1 (1, L6 IT Car3), FSTL4 (1.55, L4 IT), SNCG (2.11, L5 ET), NRP1 (1.02, Endothelial), NPY1R (1.03, L6 IT Car3)
3	16	BRINP3 (2.36, OPC), CENPF (1.93, Proliferation), LAMP5 (2.31, Lamp5 Lhx6), MEIS2 (1.88, Astrocyte), NDST4 (3.37, Sst Chodl), NTNG2 (1.44, L6 IT Car3), PAX6 (2.02, Astrocyte), PLCH1 (1.74, Sst), RFTN1 (1.83, VLMC), RORB (1.82, L4 IT), TENM1 (2.8, Chandelier), ANGPT1 (1.5, Pax6), PDGFD (1.03, Lamp5 Lhx6), ARHGAP24 (1.24, Microglia-PVM), IDH1 (1.26, Glioblastoma (Cancer cells)), CDH4 (1.13, Pax6)
4	40	AIF1 (2.68, Glioblastoma (TME)), APOE (1.01, Microglia-PVM), C1orf162 (2.1, Glioblastoma (TME)), CD14 (2.94, Glioblastoma (TME)), CD163 (4.06, Glioblastoma (TME)), CD4 (2.37, Glioblastoma (TME)), CD68 (2.7, Glioblastoma (TME)), CD86 (2.17, Microglia-PVM), CORO1A (2.25, Glioblastoma (TME)), CTSH (1.58, Microglia-PVM), CTSS (2.68, Glioblastoma (TME)), CX3CR1 (3.09, Glioblastoma (TME)), FCER1G (2.85, Glioblastoma (TME)), FCGR1A (2.4, Glioblastoma (TME)), FCGR3A (3.1, Glioblastoma (TME)), GPR34 (3.18, Glioblastoma (TME)), HLA-DMB (2.6, Glioblastoma (TME)), HLA-DQA1 (2.67, Microglia-PVM), ITGAM (2.16, Microglia-PVM), ITGAX (2.89, Microglia-PVM), ITGB2 (3.22, Glioblastoma (TME)), LRRK1 (1.65, Microglia-PVM), LY86 (1.77, Glioblastoma (TME)), LYVE1 (2.98, Microglia-PVM), MS4A6A (3.31, Glioblastoma (TME)), P2RY12 (2.83, Microglia-PVM), P2RY13 (2.52, Microglia-PVM), PTPRC (2.34, Microglia-PVM), RGS10 (2.44, Glioblastoma (TME)), RNASET2 (2.4, Glioblastoma (TME)), SPI1 (2.79, Microglia-PVM), STXBP2 (1.31, Glioblastoma (TME)), TGFB1 (2.13, Microglia-PVM), TGFBI (2.98, Glioblastoma (TME)), TMIGD3 (2.64, Glioblastoma (TME)), TREM2 (2.81, Microglia-PVM), CXCR4 (1.58, Glioblastoma (TME)), GPR183 (1.21, Glioblastoma (TME)), CAPG (1.1, Glioblastoma (TME)), GPNMB (1.1, Glioblastoma (TME))
5	33	APOE (1.55, Microglia-PVM), BCAN (3.5, Glioblastoma (Cancer cells)), BRINP3 (1.24, OPC), CALCRL (1.31, OPC), CAV1 (1.23, Glioblastoma (Cancer cells)), CSPG4 (1.91, VLMC), ERBB3 (2.67, Oligodendrocyte), GAD1 (3.57, Pvalb), GPNMB (3.22, Glioblastoma (TME)), IGFBP3 (1.15, Glioblastoma (Cancer cells)), IGFBP4 (1.06, VLMC), ITGA8 (1.23, OPC), MOG (1.71, Oligodendrocyte), NES (2.07, Astrocyte), NTNG1 (2.67, Lamp5 Lhx6), OLIG1 (3.59, OPC), OLIG2 (3.75, OPC), PCSK1 (1.24, L5 ET), PDGFRA (4.26, OPC), PLCE1 (1.61, VLMC), POSTN (3.85, L6 IT Car3), PTCHD4 (1.26, Lamp5), PTPRZ1 (2.75, OPC), RIT2 (1.06, L5 ET), SEMA5A (3.21, OPC), SERPINA3 (2.38, VLMC), SORCS1 (3.21, L5 ET), SOX10 (4.26, OPC), STK32B (3.47, OPC), TMEM132C (2.04, VLMC), TTYH1 (1.05, Glioblastoma (Cancer cells)), UGT8 (3.04, Oligodendrocyte), VCAN (4.01, OPC)
6	33	ANXA1 (1.92, Glioblastoma (Cancer cells)), CALCRL (3.47, OPC), CEMIP (3.69, VLMC), CEMIP2 (2.26, Endothelial), DCN (4.03, VLMC), FLT1 (4.14, Endothelial), IFITM3 (2.71, Microglia-PVM), IGFBP3 (3.52, Glioblastoma (Cancer cells)), IGFBP4 (3.98, VLMC), KLF2 (3.04, Glioblastoma (TME)), KLF4 (2.5, Glioblastoma (TME)), NES (1.8, Astrocyte), PECAM1 (4.15, Endothelial), PHLDB2 (2.28, VLMC), RNF144B (2.72, Endothelial), TGFBI (2.13, Glioblastoma (TME)), TGFB1 (1.78, Microglia-PVM), THSD4 (1.83, Endothelial), CAV1 (1.65, Glioblastoma (Cancer cells)), NRP1 (1.23, Endothelial), CDH6 (2.02, VLMC), HES1 (1.83, Glioblastoma (Cancer cells)), NOTCH1 (1.01, Endothelial), CD4 (1.46, Glioblastoma (TME)), LAMA2 (1.21, VLMC), THBS1 (2.18, Astrocyte), LYVE1 (1.24, Microglia-PVM), COL12A1 (1.33, VLMC), DDR2 (1.05, Pax6), SLIT3 (2.06, L5 ET), FILIP1 (1.22, L6 CT), ITGA8 (1.36, OPC), SNTB2 (1.08, L6 IT)
7	30	ADRA1A (1.32, Sncg), ALK (1.47, Chandelier), APOE (2.06, Microglia-PVM), AQP4 (2.66, Astrocyte), BCAN (1.54, Glioblastoma (Cancer cells)), COL12A1 (2.38, VLMC), CRYM (1.45, L5/6 NP), EGFR (1.23, Glioblastoma (Cancer cells)), ELOVL2 (1.87, Glioblastoma (Cancer cells)), GJA1 (4.04, Astrocyte), HES1 (1.1, Glioblastoma (Cancer cells)), HHATL (1.26, Oligodendrocyte), IDH2 (1.01, Glioblastoma (Cancer cells)), NPNT (1.42, Chandelier), OLIG1 (1.91, OPC), OLIG2 (1.65, OPC), PTCHD4 (1.33, Lamp5), PTPRZ1 (2.29, OPC), RYR3 (3.1, Astrocyte), SEMA5A (1.66, OPC), SLC17A7 (2.2, L6b), SPON1 (3.73, Astrocyte), TGFB2 (1.96, Astrocyte), TRIL (3.5, Glioblastoma (Cancer cells)), TTYH1 (3.74, Glioblastoma (Cancer cells)), TSHZ2 (1.12, L5/6 NP), CCK (1.68, Pax6), PLCE1 (1.22, VLMC), LYPD6 (1.07, Lamp5 Lhx6), EYA4 (1.05, Lamp5 Lhx6)
8	20	ANGPT1 (1.26, Pax6), BRINP3 (1.51, OPC), CHODL (2.86, Glioblastoma (Cancer cells)), CXCL14 (4.29, Pax6), CYTIP (1.8, Glioblastoma (TME)), GAD2 (1.42, Lamp5 Lhx6), LAMP5 (1.41, Lamp5 Lhx6), MEIS2 (1.1, Astrocyte), NDST4 (1.96, Sst Chodl), PAX6 (1.53, Astrocyte), RASGRP1 (1.84, L6 IT Car3), RGS16 (1.17, Microglia-PVM), SYNPR (3.38, Sst), TENM1 (1.47, Chandelier), VWC2L (2.49, L5/6 NP), MGST1 (1.15, Glioblastoma (Cancer cells)), NPY1R (1.16, L6 IT Car3), POU6F2 (1.52, L4 IT), FILIP1 (1.07, L6 CT), MAL (1.21, Oligodendrocyte)
9	19	CCL5 (5.65, Glioblastoma (TME)), CD2 (4.48, Glioblastoma (TME)), CD3G (4.2, Glioblastoma (TME)), CD48 (3.58, Glioblastoma (TME)), CD52 (5.08, Glioblastoma (TME)), GZMA (5.06, Glioblastoma (TME)), IL7R (4.04, Glioblastoma (TME)), NKG7 (4.79, Glioblastoma (TME)), KLRB1 (3.32, Glioblastoma (TME)), TRAC (2.34, Glioblastoma (TME)), THEMIS (2.63, L6 IT Car3), S100A4 (2.7, Glioblastoma (TME)), CXCR4 (1.95, Glioblastoma (TME)), CYTIP (1.48, Glioblastoma (TME)), TESPA1 (1.83, L2/3 IT), PTPRC (1.6, Microglia-PVM), GNLY (2.15, Glioblastoma (TME)), CORO1A (1.94, Glioblastoma (TME)), IDO1 (1.19, Glioblastoma (TME))
10	16	ANXA1 (2.36, Glioblastoma (Cancer cells)), APOE (2.48, Microglia-PVM), AQP4 (1.96, Astrocyte), BCAN (1.22, Glioblastoma (Cancer cells)), GJA1 (1.17, Astrocyte), IFITM3 (1.29, Microglia-PVM), MGST1 (2.65, Glioblastoma (Cancer cells)), NES (1.27, Astrocyte), SERPINA3 (3.69, VLMC), SFRP2 (2.38, Pax6), TTYH1 (1.45, Glioblastoma (Cancer cells)), VCAN (1.01, OPC), CTSH (1.74, Microglia-PVM), SPON1 (1.02, Astrocyte), THBS1 (1.38, Astrocyte), POSTN (1.37, L6 IT Car3)
11	57	KCNH5 (2.74, L6 CT), MYO5B (2.1, Pvalb), WIF1 (2.61, Pax6), ZBBX (2.87, L2/3 IT), MCTP2 (2.52, VLMC), ANKRD18A (2.59, Sst Chodl), ADAMTS12 (2.47, VLMC), IDO1 (2.56, Glioblastoma (TME)), VIP (2.47, Vip), FASLG (2.69, Microglia-PVM), OTOGL (2.6, L4 IT), NPFFR2 (2.91, L6b), LOX (2.2, Glioblastoma (Cancer cells)), MEPE (2.12, Pvalb), ANO3 (2.36, VLMC), CAV1 (2.26, Glioblastoma (Cancer cells)), TESPA1 (2.32, L2/3 IT), LYPD6B (2.15, Sncg), TPH2 (2.67, L5/6 NP), SLC17A6 (2.11, L4 IT), ADAMTS3 (2.38, L6 IT), CTNNA3 (2.44, Oligodendrocyte), ROS1 (2.13, L6b), RSPO2 (2.09, Pvalb), THEMIS (1.89, L6 IT Car3), CRHBP (2.02, Sst Chodl), CHODL (1.73, Glioblastoma (Cancer cells)), NWD2 (2.42, L6 IT Car3), POU6F2 (2.27, L4 IT), TRPC5 (1.84, L6 IT Car3), LHX6 (1.93, Chandelier), CDH12 (1.77, L6 IT Car3), SLC26A4 (1.71, L6 IT Car3), SAMD5 (2.32, Pvalb), CNTNAP3B (1.65, Chandelier), ADRA1B (2.24, Sncg), KLRB1 (1.67, Glioblastoma (TME)), NPY1R (1.61, L6 IT Car3), CD36 (1.38, L5/6 NP), OPALIN (1.71, Oligodendrocyte), ST18 (1.52, Oligodendrocyte), NXPH2 (1.55, Sst Chodl), GNLY (2.06, Glioblastoma (TME)), EYA4 (1.8, Lamp5 Lhx6), MOBP (1.3, Oligodendrocyte), HTR2A (1.67, L5 IT), NTNG1 (1.15, Lamp5 Lhx6), HTR2C (1.67, L5/6 NP), THSD4 (1.51, Endothelial), ANK1 (1.65, Chandelier), FILIP1 (1.67, L6 CT), PHLDB2 (1.49, VLMC), FSTL4 (1.47, L4 IT), SFRP2 (1.68, Pax6), VWC2L (1.4, L5/6 NP), IL7R (1.13, Glioblastoma (TME)), ADAMTS16 (2.01, L4 IT)
12	12	B4GALNT1 (1.06, Glioblastoma (Cancer cells)), CCNB2 (3.04, Proliferation), CDK1 (2.43, Proliferation), MKI67 (2.54, Proliferation), TOP2A (2.34, Proliferation), TRPC6 (1.06, Sst Chodl), NR4A2 (1.17, Glioblastoma (TME)), GAS2L3 (2.46, L6 IT Car3), CENPF (1.76, Proliferation), SDK1 (1.5, Chandelier), KLK6 (1.22, Oligodendrocyte), CCNA1 (1.21, Proliferation)
13	16	CCL4 (5.55, Microglia-PVM), CD83 (4.72, Glioblastoma (TME)), GPR183 (3, Glioblastoma (TME)), NR4A2 (1.43, Glioblastoma (TME)), ITGAX (1.68, Microglia-PVM), POSTN (1.59, L6 IT Car3), CD86 (1.73, Microglia-PVM), GPR34 (1.05, Glioblastoma (TME)), TREM2 (1.78, Microglia-PVM), FCGR1A (1.11, Glioblastoma (TME)), AIF1 (1.02, Glioblastoma (TME)), SPI1 (1.33, Microglia-PVM), CD68 (1.02, Glioblastoma (TME)), CD163 (1.29, Glioblastoma (TME)), TMIGD3 (1.08, Glioblastoma (TME)), CD14 (1.03, Glioblastoma (TME))
14	39	ADAMTS16 (3.38, L4 IT), ANO3 (2.4, VLMC), AQP4 (1.07, Astrocyte), C1QL3 (5.41, L6 IT), CCK (5.23, Pax6), CRYM (5.19, L5/6 NP), GJA1 (1.85, Astrocyte), HS3ST2 (4.22, L6 CT), PTPRZ1 (1.17, OPC), SLC17A7 (5.65, L6b), SLIT3 (4.24, L5 ET), TRIL (1.52, Glioblastoma (Cancer cells)), TSHZ2 (3.25, L5/6 NP), TTYH1 (1.71, Glioblastoma (Cancer cells)), OLIG1 (1.07, OPC), SLC17A6 (2.03, L4 IT), RXFP1 (2.87, L5 IT), SPON1 (1.44, Astrocyte), HTR2A (1.65, L5 IT), CDH12 (1.97, L6 IT Car3), ADRA1B (2.15, Sncg), PCSK1 (2.09, L5 ET), SNCG (1.64, L5 ET), COL12A1 (1.38, VLMC), RSPO2 (1.49, Pvalb), RYR3 (1.06, Astrocyte), CORO1A (1.27, Glioblastoma (TME)), CCNA1 (1.52, Proliferation), FSTL4 (1.4, L4 IT), SYNPR (1.39, Sst), TRHDE (1.98, Sst), ZDHHC23 (1.22, L6 IT Car3), NWD2 (1.41, L6 IT Car3), SORCS1 (1.05, L5 ET), RIT2 (1.2, L5 ET), HTR2C (1.35, L5/6 NP), ST18 (1.96, Oligodendrocyte), SAMD5 (1.03, Pvalb), ANK1 (1.27, Chandelier)
15	4	KCNH5 (3.38, L6 CT), NPNT (2.52, Chandelier), RORB (1.95, L4 IT), APP (1.05, Broad)
16	8	CCNB2 (4.24, Proliferation), CDK1 (4.09, Proliferation), CENPF (3.02, Proliferation), MKI67 (5.76, Proliferation), TOP2A (5.99, Proliferation), GAS2L3 (3.48, L6 IT Car3), PCNA (1, Proliferation), CCNA1 (2.07, Proliferation)
17	15	CNDP1 (6.11, Oligodendrocyte), ERMN (4.86, Oligodendrocyte), MAG (6.12, Oligodendrocyte), MOBP (3.89, Oligodendrocyte), MOG (4.91, Oligodendrocyte), CAPN3 (3.42, Oligodendrocyte), OPALIN (3.19, Oligodendrocyte), UGT8 (3.06, Oligodendrocyte), CLDN11 (2.48, Oligodendrocyte), MAL (2.66, Oligodendrocyte), ST18 (2.28, Oligodendrocyte), KLK6 (2.73, Oligodendrocyte), MYRF (2.3, Oligodendrocyte), PCSK6 (1.59, Oligodendrocyte), HHATL (1.29, Oligodendrocyte)
18	48	B4GALNT1 (1.28, Glioblastoma (Cancer cells)), MYO5B (1.7, Pvalb), NTNG2 (1.04, L6 IT Car3), FGFR2 (1, Oligodendrocyte), NR4A2 (1.01, Glioblastoma (TME)), TACR1 (2.16, Sst Chodl), SLC26A4 (2.2, L6 IT Car3), NXPH2 (2.16, Sst Chodl), LYPD6B (2.21, Sncg), SPHKAP (1.05, Lamp5 Lhx6), MEPE (1.75, Pvalb), CDH1 (2.02, Oligodendrocyte), MCTP2 (1.38, VLMC), CDH12 (1.69, L6 IT Car3), HHATL (1.58, Oligodendrocyte), KLK6 (1.95, Oligodendrocyte), WIF1 (1.38, Pax6), ANK1 (1.47, Chandelier), ANKRD18A (1.3, Sst Chodl), LHX6 (1.45, Chandelier), NWD2 (1.79, L6 IT Car3), ADRA1A (1.14, Sncg), TRPC5 (1.68, L6 IT Car3), SDK1 (1.61, Chandelier), IDO1 (1.37, Glioblastoma (TME)), SFRP2 (1.49, Pax6), MYRF (1.57, Oligodendrocyte), TPH2 (1.88, L5/6 NP), ST18 (1.36, Oligodendrocyte), ROS1 (1.69, L6b), LOX (1.3, Glioblastoma (Cancer cells)), PHLDB2 (1.1, VLMC), ZDHHC23 (1.86, L6 IT Car3), THSD7B (1.29, Sncg), FASLG (1.37, Microglia-PVM), THSD4 (1.17, Endothelial), ATP2C2 (1.77, L5/6 NP), FSTL4 (1.16, L4 IT), CTNNA3 (1.36, Oligodendrocyte), PCSK6 (1.03, Oligodendrocyte), TESPA1 (1.36, L2/3 IT), HTR2C (1.62, L5/6 NP), ADAMTS16 (1.58, L4 IT), CSPG4 (1.31, VLMC), NPFFR2 (1.23, L6b), RXFP1 (1.37, L5 IT), ZBBX (1.06, L2/3 IT), RIT2 (1.14, L5 ET)
19	18	SST (6.81, Sst Chodl), TAC1 (6.51, Sst Chodl), PVALB (5.38, Chandelier), CRHBP (3.95, Sst Chodl), LHX6 (3.62, Chandelier), GAD1 (3.23, Pvalb), ANK1 (1.79, Chandelier), ALK (2.52, Chandelier), TRHDE (2.68, Sst), SYNPR (1.31, Sst), BTBD11 (1.22, Pvalb), SNCG (2.31, L5 ET), PCSK6 (1.22, Oligodendrocyte), TRPC5 (1.75, L6 IT Car3), NXPH2 (1.62, Sst Chodl), RSPO2 (1.35, Pvalb), KIT (1.89, Proliferation), SAMD5 (1.36, Pvalb)
20	47	THBS1 (5.73, Astrocyte), SLC17A6 (3.97, L4 IT), PCSK1 (4.39, L5 ET), ALK (4.82, Chandelier), GAD1 (4.07, Pvalb), ANK1 (2.65, Chandelier), CALCRL (3.42, OPC), PLCE1 (3.68, VLMC), CORO1A (2.57, Glioblastoma (TME)), NPY1R (2.25, L6 IT Car3), CNTNAP3B (2.3, Chandelier), RIT2 (2.23, L5 ET), ADRA1A (2, Sncg), SAMD5 (2.43, Pvalb), CCNA1 (2.85, Proliferation), LYPD6 (1.9, Lamp5 Lhx6), CX3CR1 (1.75, Glioblastoma (TME)), NES (1.92, Astrocyte), CDH4 (1.75, Pax6), OLIG2 (1.28, OPC), IGFBP3 (2.06, Glioblastoma (Cancer cells)), PTPRZ1 (1.54, OPC), NTNG1 (2.11, Lamp5 Lhx6), TPH2 (1.46, L5/6 NP), AIF1 (1.61, Glioblastoma (TME)), SERPINA3 (1.15, VLMC), CD83 (1.17, Glioblastoma (TME)), FCGR3A (1.49, Glioblastoma (TME)), STAT3 (1.21, Endothelial), SNTB2 (1.38, L6 IT), ANXA1 (1.01, Glioblastoma (Cancer cells)), BCAN (1.28, Glioblastoma (Cancer cells)), P2RY12 (1.02, Microglia-PVM), HLA-DQA1 (1.17, Microglia-PVM), TREM2 (1.73, Microglia-PVM), SOX10 (1.34, OPC), TTYH1 (1.1, Glioblastoma (Cancer cells)), PSEN2 (1.08, Glioblastoma (Cancer cells)), HLA-DMB (1.15, Glioblastoma (TME)), ZDHHC23 (1.18, L6 IT Car3), SORCS1 (1.85, L5 ET), PTCHD4 (1.09, Lamp5), STK32B (1.95, OPC), LRRK1 (1.2, Microglia-PVM), LY86 (1.09, Glioblastoma (TME)), CCK (1.49, Pax6), SNCG (2.48, L5 ET)
21	50	GPNMB (5.89, Glioblastoma (TME)), CD68 (2.88, Glioblastoma (TME)), TREM2 (3.13, Microglia-PVM), CTSS (3.09, Glioblastoma (TME)), PDGFRA (2.99, OPC), HLA-DMB (2.53, Glioblastoma (TME)), THBS1 (3.48, Astrocyte), MS4A6A (2.26, Glioblastoma (TME)), ANXA1 (2.68, Glioblastoma (Cancer cells)), APOE (2.21, Microglia-PVM), CD163 (2.43, Glioblastoma (TME)), CD48 (3.93, Glioblastoma (TME)), CEMIP (4.17, VLMC), CTSH (2.44, Microglia-PVM), FCER1G (1.71, Glioblastoma (TME)), HS3ST2 (3.7, L6 CT), CD4 (2.6, Glioblastoma (TME)), GPR34 (2.3, Glioblastoma (TME)), SOX10 (2.59, OPC), ITGAM (2.35, Microglia-PVM), AIF1 (2.22, Glioblastoma (TME)), VCAN (2.15, OPC), ITGB2 (2.96, Glioblastoma (TME)), IFITM3 (1.98, Microglia-PVM), CXCR4 (2.23, Glioblastoma (TME)), CAPG (2.84, Glioblastoma (TME)), ERBB3 (2.07, Oligodendrocyte), CD14 (2.47, Glioblastoma (TME)), IGFBP4 (2.84, VLMC), P2RY13 (1.94, Microglia-PVM), CD86 (1.48, Microglia-PVM), UGT8 (1.35, Oligodendrocyte), SORCS1 (2.87, L5 ET), SEMA5A (1.61, OPC), PLCE1 (1.02, VLMC), GAS2L3 (1.6, L6 IT Car3), LYVE1 (1.81, Microglia-PVM), NES (1.32, Astrocyte), POSTN (1.18, L6 IT Car3), DCN (2.24, VLMC), OLIG2 (1.52, OPC), HLA-DQA1 (2.32, Microglia-PVM), LY86 (1.09, Glioblastoma (TME)), PTPRC (1.82, Microglia-PVM), SERPINA3 (1.23, VLMC), CDH6 (2.09, VLMC), GPR183 (2.84, Glioblastoma (TME)), P2RY12 (1.93, Microglia-PVM), TGFBI (2.4, Glioblastoma (TME)), CCL4 (2.34, Microglia-PVM)

Preliminary annotation

With the information above, we can begin to piece together some preliminary definition for each cluster.

This Cell Paper and this Nature Paper

prelim_annotation_res1 = data.table(t(setDT(list(
    `1` = c('Neuron','GABAergic Neuron (lamp5, pax6)'),
    `2` = c('Neuron','Glutamatergic Neuron (GBM CC Xenium)'),
    `12` = c('Neuron','GABAergic Neuron (Proliferating)'),
    `15` = c('Neuron','Glutamatergic Neuron'),
    `18` = c('Neuron','Unknown Neuron (GABAergic and Glutamatergic)'),
    `17` = c('Oligodendrocytes','Oligodendrocytes'),
    `16` = c('Proliferation','Proliferation'),
    `9` = c('Lymphoid','Lymphoid'),
    `11` = c('Neuron','Unknown Neuron (GABAergic and Glutamatergic)'),
    `19` = c('Neuron','GABAergic Neuron (sst, sst/chodl, pvalb, chandelier)'),
    `3` = c('RGC','RGC'),
    `8` = c('RGC','RGC'),
    `14` = c('Neuron','Glutamatergic Neuron (l6b)'),
    `20` = c('Undetermined','Undetermined'),
    `4` = c('Myeloid','Myeloid'),
    `6` = c('Stromal','Endothelial/VLMC'), 
    `13` = c('Myeloid','Myeloid'),
    `5` = c('OPC','OPC'),
    `7` = c('Astrocyte','Astrocyte'),
    `10` = c('Astrocyte','Astrocyte (VLMC, myeloid)')
))), keep.rownames = 'cluster')[
    , c('Coarse Cell Type', 'Fine Cell Type Annotation') := list(V1, V2)
][,c('V1','V2') := NULL][
    order(as.numeric(cluster)), 
    `Coarse Cell Type UMAP` := paste0(
        `Coarse Cell Type`, ' ', c(1:.N)
    ),.(`Coarse Cell Type`)
][,`Fine Cell Type` := gsub(
        ' \\(.*','', `Fine Cell Type Annotation`
)][
    order(as.numeric(cluster)),
    `Fine Cell Type UMAP` := paste0(
        `Fine Cell Type`, ' ', c(1:.N)
    ),.(`Fine Cell Type`)
]
setcolorder(
    prelim_annotation_res1,
    c(
        'cluster','Coarse Cell Type','Coarse Cell Type UMAP',
        'Fine Cell Type','Fine Cell Type UMAP',
        'Fine Cell Type Annotation'
    )
)
kable(prelim_annotation_res1)
write.table(
    prelim_annotation_res1,
    file = paste0(
        workdir,'coja/Spatial_PLGG/data/Xenium/banksy/banksy_cell_cluster_annotation.tsv'
    ),
    sep = '\t', quote = FALSE, row.names = FALSE
)

cluster	Coarse Cell Type	Coarse Cell Type UMAP	Fine Cell Type	Fine Cell Type UMAP	Fine Cell Type Annotation
1	Neuron	Neuron 1	GABAergic Neuron	GABAergic Neuron 1	GABAergic Neuron (lamp5, pax6)
2	Neuron	Neuron 2	Glutamatergic Neuron	Glutamatergic Neuron 1	Glutamatergic Neuron (GBM CC Xenium)
12	Neuron	Neuron 4	GABAergic Neuron	GABAergic Neuron 2	GABAergic Neuron (Proliferating)
15	Neuron	Neuron 6	Glutamatergic Neuron	Glutamatergic Neuron 3	Glutamatergic Neuron
18	Neuron	Neuron 7	Unknown Neuron	Unknown Neuron 2	Unknown Neuron (GABAergic and Glutamatergic)
17	Oligodendrocytes	Oligodendrocytes 1	Oligodendrocytes	Oligodendrocytes 1	Oligodendrocytes
16	Proliferation	Proliferation 1	Proliferation	Proliferation 1	Proliferation
9	Lymphoid	Lymphoid 1	Lymphoid	Lymphoid 1	Lymphoid
11	Neuron	Neuron 3	Unknown Neuron	Unknown Neuron 1	Unknown Neuron (GABAergic and Glutamatergic)
19	Neuron	Neuron 8	GABAergic Neuron	GABAergic Neuron 3	GABAergic Neuron (sst, sst/chodl, pvalb, chandelier)
3	RGC	RGC 1	RGC	RGC 1	RGC
8	RGC	RGC 2	RGC	RGC 2	RGC
14	Neuron	Neuron 5	Glutamatergic Neuron	Glutamatergic Neuron 2	Glutamatergic Neuron (l6b)
20	Undetermined	Undetermined 1	Undetermined	Undetermined 1	Undetermined
4	Myeloid	Myeloid 1	Myeloid	Myeloid 1	Myeloid
6	Stromal	Stromal 1	Endothelial/VLMC	Endothelial/VLMC 1	Endothelial/VLMC
13	Myeloid	Myeloid 2	Myeloid	Myeloid 2	Myeloid
5	OPC	OPC 1	OPC	OPC 1	OPC
7	Astrocyte	Astrocyte 1	Astrocyte	Astrocyte 1	Astrocyte
10	Astrocyte	Astrocyte 2	Astrocyte	Astrocyte 2	Astrocyte (VLMC, myeloid)